Skip to content

Fastest and only macOS Dictation app with on-device STT and custom trained AI enhancement model. Windows, iOS testflight! Android, Linux waitlist open! A local Wispr Flow alternative. Follow https://x.com/fluidvoiceapp for more

fix: require complete speaker turn transcription

postoso54f6e85b20cb
diff --git a/Sources/Fluid/Services/MeetingTranscriptionService.swift b/Sources/Fluid/Services/MeetingTranscriptionService.swift
index d25d9cfa..8d046460 100644
--- a/Sources/Fluid/Services/MeetingTranscriptionService.swift
+++ b/Sources/Fluid/Services/MeetingTranscriptionService.swift
@@ -659,11 +659,11 @@ final class MeetingTranscriptionService: ObservableObject {
minimumDurationSeconds: 1.1
)
// Mirror the standard path's 1-second ASR minimum.
- guard samples.count >= 16_000 else { continue }
+ guard samples.count >= 16_000 else { return nil }
let chunkResult = try await provider.transcribe(samples)
let text = chunkResult.text.trimmingCharacters(in: .whitespacesAndNewlines)
- guard !text.isEmpty else { continue }
+ guard !text.isEmpty else { return nil }
pieces.append(text)
confidenceSum += chunkResult.confidence